Enable torch.xpu._XpuDeviceProperties in Helion kernel

[EikanWang]

Currently, the grouped_gemm_jagged_persistent in examples/grouped_gemm.py gets the number of workers by querying multi_processor_count of device properties for CUDA. XPU provides a similar API. This PR intends to extend the support to XPU.

[Copilot]

Pull Request Overview

This PR extends XPU support in the Helion kernel by enabling torch.xpu._XpuDeviceProperties to work with the grouped GEMM functionality. The changes allow the kernel to query XPU device properties (specifically gpu_subslice_count) similar to how it currently queries CUDA device properties (multi_processor_count).

Key changes:

• Added XPU device property support in the type propagation system
• Updated grouped GEMM example to handle both CUDA and XPU devices
• Modified test expectations to reflect the new conditional logic

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
examples/grouped_gemm.py	Added XPU device detection and property querying logic, updated main function to prefer XPU when available
helion/_compiler/type_propagation.py	Extended type propagation to support torch.xpu._XpuDeviceProperties alongside CUDA properties
test/test_examples.expected	Updated test expectations to match the new conditional device property logic

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Corrected grammar in comment: 'we change update it' should be 'we should update it'.

Suggested change

	# TODO(EikanWang): gpu_subslice_count is an out-of-date term. we change update it to XeCore number.
	# TODO(EikanWang): gpu_subslice_count is an out-of-date term. we should update it to XeCore number.

[oulgen]

Suggested change

	if type(value) in [
	torch.cuda._CudaDeviceProperties,
	torch.xpu._XpuDeviceProperties,
	]:
	if isinstance(value, (torch.cuda._CudaDeviceProperties,
	torch.xpu._XpuDeviceProperties)):

why not this?

[EikanWang]

Sure.